Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
Patch: make adaptive routing model-agnostic (no Claude requirement)
What This Skill Does
Interviews users about their profession, tools, and pain points, then designs and builds a team of 3-7 specialized AI agent personas. Each agent gets a distinct personality, a defined role, inter-agent coordination rules, and a self-improvement logging system. Adaptive model routing (Fast/Think/Deep/Strategic) is configured across the council.
It runs structured discovery before building anything, so agents reflect the user's actual daily workflow rather than generic presets.
When to use it
- Building a dedicated research and writing agent team for a content creator
- Setting up a code review and deployment council for a solo developer
- Creating finance, scheduling, and outreach agents for a small business owner
- Organizing a multi-agent workflow for recurring client deliverables
- Spinning up specialized agents after analyzing an existing OpenClaw workspace
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Build me an agent team for my dev workflow
- 1Asks about profession, tools, and daily platforms in a first batch of questions
- 2Follows up on time-consuming tasks and pain points in a second batch
- 3Designs a council of 3-5 agents with names, roles, and a coordination map, then presents a table for user approval
- 4Runs init-council.sh to create the directory skeleton, then populates SOUL.md, AGENTS.md, and .learnings/ files for each agent
- 5Sets up adaptive model routing thresholds in root AGENTS.md and generates an architecture doc
A fully configured agent council in the workspace with per-agent personalities, coordination rules, self-improvement logs, and adaptive routing